Robust Protection of Networks against Cascading Phenomena

Recently, the spread of many different undesirable phenomena has been modeled as a network diffusion process cascade, e.g. misinformation spread in social networks, dissemination of infectious disease through different regions, virus and malware propagation through computer networks, expansion of fire in forests and towns, spread of invasive species across landscape etc. The consequences of such spreads can be severe and hence it is important to design and develop efficient ways of protection. In this thesis, we address the problem of robust network protection against the spread of such cascades. Particularly, we look at the robust protection against a set of the most undesirable cascades, where our goal is to block the smallest possible set of nodes that is robust against the worst possible spread of these cascades. We propose an efficient approximation algorithm that based on a greedy strategy, finds a near-optimal solution for a given set of cascades. Moreover, we design an efficient iterative procedure that incorporates this algorithm. The procedure allows the greedy algorithm to find a near-optimal solution by considering a relatively small number of the given cascades. We also look at different protection actions to which we refer as blocking mechanisms. The types of blocking mechanisms that we consider here are “static”, i.e. not-spreading, and “dynamic”, where once a node is blocked it influences other nodes to become “blocking nodes” as well. We use large real-world networks to evaluate our procedure. Empirical results show that our procedure significantly outperforms well-known heuristics for all considered blocking mechanisms.